The Setting
A finite address carries both contraction data and translation randomness. Conditioning on block types separates these two features in the finite translation law and in its affine mixture with a future measure.
A weighted affine iterated function system on a finite alphabet is a family of similarities with strictly positive normalized weights , as in (1). (1)✦ For a word , its complete affine map is the pair of signed slope and translation, its word similarity is , and its word probability is , defined by (2). (2)
For positive weights on a finite set , a tag and points , the tag fiber is the set of points with the same tag as , its mass is , and the posterior is the conditional probability of given that tag, as in (3). (3) For positive weights on a finite set , a tag and a point , the conditional observable distribution of an observable given the tag of is the distribution of under the weights in (3).✦
A finite probability law on is a finite support with strictly positive point weights summing to one.✦ For strictly positive weights summing to one on a finite set and an observable , the observable law of is the finitely supported probability law on its image whose weight at is the fiber mass in (4).✦ The Shannon entropy of the weights in nats and their observable entropy are also given there, with . (4) The Shannon entropy in bits of a finite probability law on is the real number obtained by dividing its point entropy by .✦ For a finite probability law on and a scale , the grid entropy is the entropy in bits of the grid-cell observable of , as in (5). (5)
The Claim
Definition 1 (Block types and conditional translation laws). Let the system be as in (1), and let be an integer. The -block system is the weighted affine system on whose map at a block is and whose weight is .✦ The word type of a block in the -block system is its vector of symbol counts, defined by (6). (6)✦ For a word of length , its block type sequence is the sequence of count vectors in its consecutive blocks, given by (7). (7)✦ For two blocks with the same word type, their signed ratios satisfy (8). (8)✦
Let be any tag, and let be the tag sequence of a word.✦ For words and a tag , the conditional word probability is the product of the symbol posteriors from (3), with , as in (9). (9)✦ For a tag and a word , the conditional translation law is the distribution of the translation conditioned on the tag sequence , as defined by (10). (10)✦ For a tag , a word and a Borel measure on , the conditional affine tail is the Borel measure given by the mixture in (11), where denotes pushforward; the future measure remains unconditioned. (11)✦ These definitions also apply to the block system, with its word type as the tag.
The contraction depth of a symbol is the real number given by (12). (12)✦ For a word over , its contraction depth is given by (13). (13)✦ For real parameters and an integer , the fine scale is the positive real number in (14), and the band start associated with this scale is the integer in (15). (14)✦ (15)✦ For a tag on , a word over and a scale , the conditional entropy loss is the real number measured in nats given by (16). (16)✦ For a tag on , real parameters and an integer , the retained mean entropy loss is the real number obtained by summing the weighted losses over addresses of depth at most , as in (17). (17)✦
In the paper, the result reads: for the self-similar probability measure , with lower Hausdorff dimension , base-two map-walk entropy rate and positive base-two Lyapunov exponent satisfying , choose with and an integer with , where , and set , and , so that and ; a -letter block of independent and identically distributed symbols has count type , signed ratio and uniform conditional law on words of that type; independent types have their induced law , and, writing and , a bound on one block’s depth gives ; the first types generate , and conditional on the entire environment , independent choices with laws define as the law of and as the corresponding infinite sum law, whose type average is ; the scale and depth event are , and .
What Is Checked
The constructions are stated over finite weighted affine systems, arbitrary tags on their alphabets and Borel future measures, with real scale parameters. Every nonempty word has signed slope of modulus less than one✦; at each fixed length the word probabilities are strictly positive✦ and sum to one.✦ The normalized weights on every tag fiber sum to one.✦